Saturday November 15, 2008

Kelly Gaither, Associate Director at TACC – University Texas Austin, talked today about Spur, a new scalable visualization system that is directly connected to Ranger, the 500 TFLOP Sun Constellation System at the Texas Advanced Computing Center (TACC) in Austin.

Spur, which was a joint collaboration between TACC and Sun, bolts the visualization system directly into the InfiniBand fat tree used to create the 65K-core Ranger cluster. This allows both compute and visualization to be used effectively together, essential for truly interactive data explorations.

Spur is actually a cluster of eight Sun compute nodes, each with four nVidia QuadroPlex GPUs installed. The overall system includes 32 GPUs, 128 cores, close to 1 TB RAM and can support up to 128 simultaneous Shared Visualization clients. Sun Grid Engine is used to schedule jobs onto Spur nodes.

Spur went into production in October and within a week was being used about 120 hours per week for interactive HPC visualization. Users who access Spur via the TeraGrid find the resource invaluable because the alternative--computing on Ranger and then transferring results back to their home site to visualize it just isn't feasible due to the sheer volume of data involved. One user estimated it would take him a week to transfer the data to his local facility whereas with Spur he is able to compute at TACC, render the visualization at TACC using Spur, and then use Sun Shared Visualization software to display the visualization at his local site.

Prof. James Leylek – Executive Director, CU-CCMS at Clemson spoke today at the HPC Consortium Meeting in Austin about his experiences with the installation and acceptance of their new Sun Sun Blade 6000 system. CU-CCMS installed 31 SunBlade 6000 chassis with 10 blades per chassis each with two Intel quad-core CPUs for a total of 3440 cores, connected with DDR InfiniBand. The entire cluster has 14 TB RAM and delivers about 34.4 TFLOPs peak performance. After deciding on the system, the next major task was to specify an acceptance test for the cluster.

Initially, CU-CCMS decided that an uninterrupted 72-hour LINPACK run would be an appropriate full-system acceptance test. External advisors, however, suggested that such a long run on a system of this size would be infeasible. As it turns out, the full-scale LINPACK ran to 48 hours without any problems. And then to 72 hours. And then to 130 hours.

The entire project was completed in two weeks and two days with two local resources assigned to the acceptance process. End result? A very happy HPC customer. We like that.

I arrived last night in Austin for the Sun HPC Consortium meeting this weekend and for Supercomputing '08 next week. I joined several colleagues for a casual dinner at a local home hosted by Deirdré and her daughter Ross and attended by several of Ross' friends. It was a fun and relaxing way to ease into this trip, which year to year proves to be pretty exhausting since the Consortium runs all weekend and is followed immediately by Supercomputing, which we can count on to deliver a week of high-energy, non-stop sensory overload.

Thanks to Deirdré for the invitation and to Ross, Mo, Trishna, April, Griffin, and Terry for the entertaining evening and the extremely wide-ranging conversation (wow!) And a special thanks to Ross for hosting and for introducing me to a Sicilian pesto that's to die for! YUM. In return, I hope they enjoyed my "opening the apple" demo. :-)

Our internal training session for Sun's HPC field experts (HPC ACES) will finish in about an hour, at which point we will break for lunch and the return for the start of the HPC Consortium meeting.

Let the games begin!